【作者】 王永刚；

【导师】 胡涛；

【作者基本信息】 西安理工大学 ， 信号与信息处理， 2002， 硕士

【摘要】 彩色印刷复制过程是颜色的分解、转换、传播和再现的过程，其目的则是尽量忠实地再现原稿。为获得高质量的彩色印品，检测和控制各色网点的变化是非常关键的。 针对当前印刷领域中印刷图像色彩检测和控制的方式及设备存在的不足，本课题提出了一种新的基于数字图像处理和模式识别技术的色空间转换及分色原理，同时设计出了基于色度检测法的网点检测和识别系统——DIA (Dot Image Analyzer)系统。DIA系统借助CCD数码相机和光学显微镜等设备直接进行像内检测，获取显微网点图像的RGB数据信息；而后采用数字图像处理和模式识别技术对网点图像中各像素进行颜色模式识别(RGB→CMYK)；最终得到各色网点图像的分色结果以及各色网点面积率等网点参数，这些参数可用于印刷机上各色各墨区的供墨量调节与控制。DIA系统中算法模型的设计基于“分而治之”和“由粗到精”的模式识别思想，具体构建了两种粗分器(BDT和CMAC)及三种精分器(CLC-FPP、CPN-WED和NNE-IFM)，这些分类器的设计综合了聚类法、决策树、神经网络、模糊逻辑以及信息融合等技术的应用，并使其优势互补，因此在理论上具有识别精度高及速度快的优点。课题对各种分类器的构建过程及工作原理进行了细致准确的论述，并以Microsoft Window Me为操作平台采用Microsoft Visual Studio C++6.0为工具开发出用于网点图像分色处理的DIA系统配套软件，该软件同时还具有相应的图像处理功能。最后对算法模型中的各种分类器进 西安理工大学硕士学住论文行了实验验证和误差分析，同时指出了后续研究的方向。 本课题的研究对进一步开发现代印刷机的闭环色彩控制有直接的指导作用：课题中提出的色空间转换及分色原理对深入研究色彩复制理论有重要的理论价值；开发出的软件在印刷质量检测和控制以及图像处理中有一定的应用价值。更多还原

【Abstract】 Color Printing is a process of decomposing, transforming, transmitting and reproducing color information, and its purpose is to copy originals faithfully. To get the print with high quality, it is a key to inspect and control the varying trend of dots of each color class.Due to the drawback of methods and devices of inspecting and controlling printing images in current printing industry, a new principle for transforming color spaces and color separation based on digital image processing and pattern recognition is presented, and a system for inspecting and recognizing dots ?DIA (Dot Image Analyzer) system based on colorimetric measurement is designed. With CCD digital camera and optical microscope, DIA carries out measurements directly within images and obtain its RGB information; then recognizes the color pattern of each pixel in micro images (i. e. RGB-CMYK); at last color separating results and some parameters of dots of each color class, which can be used to adjust and control ink amounts of each ink zone of each color on printers. On the basis of the concepts of Divide-and-Conquer and Rough-Fine, DIA build a set of algorithm models including two rough classifiers (i.e. BDT and CMAC) and three fine classifiers (i. e. CLCHFPP, CPN-WED, NNE-IFM) which integrate some techniques such as clustering, decision tree, neural networks, fuzzy logic and information fusion etc. and take advantage of them, so DIA has a high accuracy and speed theoretically. The thesis discussed the building processes and working principles of presented classifiers, and developed the corresponding software for color separation of dot images with the operation system of Microsoft?Window* Me and Microsoft Visual Studio C++ 6. 0. In the end, some experiments were carried out to verify these classifiers and error analysis was given; the next research direction was pointed out.The thesis will directly help to further developing the close looped color control system of modern printers. The principle of transforming color spaces and color separation greatly contributes to the deep research of color reproduction in theory and the software developed has some applied value in printing quality control and image processing.更多还原